Automatic Extractive Text Summarization for Sinhala Language using Textrank, TF-IDF and Fuzzy Clustering Algorithms

This paper studies how modern text summarization methods can be used for the Sinhala language. It uses techniques like Fuzzy Clustering, Text Ranking, and Term Frequency-Inverse Document Frequency (TF-IDF) to shorten the text. The study focuses on creating brief summaries from a chosen set of Sinhala paragraphs in a specific area. Python was employed to implement these techniques, with a manually curated list of stop words incorporated to minimize redundancy and enhance the summarization quality. The generated summaries were then compared with a manually produced summary to evaluate the effectiveness of the automated methods. This comparison highlights the strengths and limitations of each technique in handling the unique linguistic characteristics of Sinhala, providing valuable insights into the development of more effective summarization tools for underrepresented languages. The average F1 score obtained for the TextRank method was 58.33%, while the Fuzzy Clustering method achieved 18.52%, and the TF-IDF method achieved an F1 score of 8.33%. This indicates that the Fuzzy Clustering approach outperforms TF-IDF, and TextRank outperforms both Fuzzy Clustering and TF-IDF in the context of summarizing Sinhala text.